[lavishlatern]

I don't see how replacing all UTC in software with TAI is more realistic than breaking UTC sync with UT1 (isn't it literally doing the same thing?). The whole point is that going forward, leap seconds are going to get harder to deal with. Especially in the case of a negative leap second, which seems like a more "true" y2k-like scenario.

[Veserv]

The difference is that replacing usage of UTC with TAI is a voluntary choice made for each program, but redefining UTC to be a fixed offset relative to TAI, which is effectively just redefining UTC to be TAI, is a forced change on everything everywhere all at once that everybody has to handle because one of their dependencies changed.

It would be like silently changing the start of unix epoch time to 1800 instead of adding a new “Unix time since 1800” and asking people to switch.

[lavishlatern]

Not at all. Everybody using UTC would just not need to deal with leap seconds anymore. A UTC second is the same as a TAI second. It's a no-op for the vast majority of UTC users. UTC will just drift slightly more from UT1.

This change only affects people who need UTC to be close to UT1 and also somehow don't know what UT1 is.

[Veserv]

Sure, everybody using UTC when they actually want TAI would be a no-op, but then you irreversibly break everybody who actually wants UTC and assumed that UTC would not change meanings.

The people who would be unaffected by the redefinition can already just trivially switch manually (as we already assumed that just redefining things under them would work), leaving the UTC people alone. There is no good reason to silently break all programs carefully designed to use UTC correctly to fix all of the programs haphazardly written by people who did not know what they were doing and used UTC when they actually wanted TAI. Especially since fixing the wrong use of UTC is so trivial that we assume it can be done with no modification.

[redcitylights]

‘Programs carefully designed to use UTC’ would only irreversibly break by very slowly becoming out of sync with the rotation of the earth.

A few applications should switch standards, the question is whether solar concerned applications should switch to UT1, or continuity concerned applications should switch to TAI. The former is simpler, easier, cheaper, and only causes unexpected behavior (quite slowly), NOT systematic failure.